The use of AI in interviews is going to make things much worse for candidates initially, but if done right, much better for all parties in the long run.
This is the second article in a three-part series on AI’s impact on the hiring process. The first article focused on how the resume itself will evolve given the incredible text processing capabilities of LLMs. In this second article, we see how the same technology will change the dynamics of interviewing. The series concludes by understanding how these two changes, along with other newly unlocked capabilities will alter the hiring process. While there are potential issues for bias, if done right, AI can make the entire job market much more efficient benefiting candidates and companies alike.
Given the nature of my work, both as someone who hires people at my companies and the work I do with The Career Toolkit: Essential Skills for Success That No One Taught You training companies and candidates, I regularly speak with people in HR and recruiting and watch for trends. We are on the cusp of a revolution in interviewing: the AI interview. You may have read my previous article, ”How Companies Can Use AI to Make Their Interview Process More Effective”, on using AI to generate good questions. You likely also know that AI is sometimes used by ATSs (Application Tracking Systems) to filter candidates. That’s just the tip of the iceberg. We’re starting to see the use of AI as the first interviewer.
This is actually not completely new. In 2012 the company I worked for subleased some office space to a recruiting agency that used AI (although not an LLM) in interviews. The human interviewer would speak with the candidate by phone, and the conversation was recorded and passed to a specialized AI tool that would provide an analysis of the candidate. They focused specifically on sales, so were specialized in terms of what the AI’s assessment would focus on.
Video interviews are also not new. We’ve had the technology for years to provide a set of prompts and have the candidates record audio or video responses. I’ve seen a handful of companies use it as a first pass. Historically those recordings, usually of 30-90 seconds, were reviewed by humans.
With today’s LLMs the human can be removed from the loop. Those recorded answers in video or audio format (or text answers in an automated chat session) can be parsed by LLMs to select candidates. Very soon, if not today, an LLM can even dynamically conduct the interview, responding to prior answers and creating follow-up questions as would a human. It may not be commonplace yet, but we can see it coming.
But there’s a catch. When a candidate shows up to an interview the candidate spends an hour of his time interviewing with the company and an employee spends an hour of her time interviewing the candidate. It’s balanced. (We make the candidate incur the cost of traveling to the office typically but that’s more for logistical reasons.) When ten candidates each interview for an hour with the company, the company has spent an equivalent ten hours interviewing candidates. The company is investing as much time as the candidate in the process. If we truly want interviews to be a two-way street, this seems appropriate.
Now imagine if a company says, “Spend thirty minutes recording answers to these questions, which we’ll feed to an LLM to do our first round of interviewing.” Companies may ask this of hundreds of candidates who cumulatively invest hundreds of candidate hours while the company invests nothing (other than CPU cycles). It’s asymmetric. While asymmetry itself isn’t wrong, it means candidates are going to need to jump through many hoops and the company does not. Companies will put up more hoops (especially in employer-friendly job markets) and candidates will get even more frustrated, investing more time upfront than they have in the past. While there are challenges making the job application process too easy as noted in Part 1, it’s also not good to make it too cumbersome on the candidate.
But we also have the answer. Long ago we created a standard document which can be reused for any job application or reused with minor changes: the resume. Candidates don’t need to start from scratch each time they apply to a job, they can take the same resume and re-use it over and over. We haven’t done that for interviews because until the past few decades we didn’t have an economically feasible way to record, share, and analyze those responses. Today we do.
At least half the questions you get in an interview are standard; likely more than half in many cases. Why are you looking for a new role? What did you achieve at company X? How do you handle a difficult customer? What was your most successful marketing campaign? How do you approach hiring your team? What’s your management style? What’s your biggest weakness? Some are general, while others are specific for a given role, but the point is you answer some version of them repeatedly across your many interviews during a job search process.
What if you answered them once, and companies could parse your answers without you needing to be there? Just like you wrote your resume once and companies can look at it on their time, they can do the same with your answers. No human will have the time to listen to a dozen or more answers from every applicant, let alone every candidate who matches a role on a job board search; but they don’t have to, AI agents can do it for us.
I’m not saying this replaces the interview. There will absolutely still need to be in-person interviews, but now we can do a first pass of companies finding or selecting candidates not just from a resume, but from more detailed responses historically not available until after an interview has happened. In, “The Future of the Resume, AI Interviews, and the Evolution of Hiring – Part 1: Less is More is Now More is More” we saw that more information can be used for matching now that AI is being used and proposed the concept of an enhanced resume. Here, too, we can get even more information, now in audio, video, or text, and that can be used asynchronously, a prior, to see if it’s worth pursuing (on both sides, as we’ll get to below). Think of it as an "enhanced interview” that begins by using pre-recorded answers and candidate statements.
Even better, we can now match on deeper issues. As I noted in, “The Streetlamp Effect in Hiring”, hiring teams often don’t know how to assess soft skills and cultural fit. I believe AI can be better at that than many humans. Of course, this is because the bar for human interviewers is so low it’s basically sitting on the ground. I don’t think today's AI is better than the best human interviewers, and it may never be. Even so, AI assisting humans in this can make even mediocre humans much better at it. That’s better for the companies and better for the candidates.
AI can even go further and incorporate feedback from internal assessments of what qualities make employees successful in certain departments and include that as part of its filtering process. Something most humans don’t even know about, let alone do with any competence. (We’ll discuss this more in Part 3.)
And whereas keyword searches on resumes are binary or purely quantitative (e.g., this resume had a 74% match on keywords for the role), LLM processing of the extended prose in answers to interview questions can be more nuanced. A good AI agent might note that a candidate is missing a qualification or two (not uncommon when companies search for a unicorn in the middle of an overly complicated Venn diagram of a dozen skills) but is strong in other aspects that make the candidate worth considering. It benefits both candidates and companies to evaluate people with a wider lens. That’s been costly to date, but LLM text processing changes the cost-structure to make it feasible.
This won’t be the only change. Better hiring teams and candidates know that interviews are a two-way street, and the above sounds more like it’s only in one direction. (Before you have a Heart Attack, or feel you have No Control, I promise that One Way or Another this will be addressed, even if not Right Now, at the start of the change.) Companies, too, can prerecord answers.
Already some companies talk about corporate culture on their website. They can continue to use text, as well as offer audio or videos with more information about the company. Likewise, the hiring manager can also pre-record some answers. If you’re not sure what answers should be pre-recorded, check out my list of Candidate Interview Questions which covers topics like corporate culture, management style, support, and job details. These are questions you should be getting asked by each candidate making the process tedious for you. Again, a “write-once, read-many” solution can be used. If you’re a manager hiring for multiple roles, as with the candidates, one recording can be reused across roles. This will save time for the hiring team (not having to repeat so many answers). It also means we can now have a candidate’s AI agent evaluate the company in parallel with the company’s AI agent evaluating the candidate.
Some readers will note that this will help, and hurt, certain candidates. For example, some people are good at public speaking (and interviewing is a specialized form of public speaking), and others are not. Some people are good at interviewing, others are not. Some candidates will now do much better because instead of needing to respond to a question within a few seconds when asked in person, they can plan and prepare a recorded answer without time pressure and can try a few times to get it right. While for salespeople or executives answering questions on the spot may be a key attribute (and so should be assessed as part of the interview process), for many jobs it is not. On the other hand, some people are nervous about being recorded, and now there will be bias against them in this step of the process. No system is perfect. At least for those who struggle with being recorded, they can get coaching and try over and over when recording answers. The nervous interviewer can’t bring a coach to the interview or try an answer four or five times until he gets it right.
As in Part 1, a word of warning. Bias is well established in AI training data. Legacy bias in hiring will translate to bias in AI reviews of resumes, interviews, and other parts of the application process; see, “Redlining in the Twenty-First Century: Everything Everywhere All at Once.” For this system to work well, we need to address the issue of bias.
Again, AI won’t completely replace the human interview. Years ago, some companies started with a phone screen to quickly filter candidates before asking both the candidate and interviewers to invest a full hour of their time. This is just the next logical step in that process. The “phone screen” is now both the company’s AI agent and the candidate’s AI agent reviewing the more detailed information on the company to see if there’s a fit and only then will both sides invest their time in an in-person interview. This saves time for everyone. It also helps to create a more consistent first filter for candidates (bias aside).
It should be noted that even the in-person interviews may be AI-assisted. The interview may be recorded and assessed by AI off-line. Additionally or alternatively, AI may listen in and help the interview focus on key areas in the candidate. Importantly, if a company should do this, then I would strongly propose party parity. If the company is recording the interview the candidate should be able to get a copy of that recording so she can review it offline as well, both to improve herself and to assess the company’s answers to her questions. If the company is having AI listening in to help guide the conversation the candidate should also have his AI agent listen in. This can be both to guide the candidates’ questions at the end, but also to “coach” the candidate. For example, if the candidate answers a question but forgets to mention a key accomplishment, his AI agent can “nudge” the candidate to mention it. After all, aren’t both the company and candidate better off with a more complete picture?
As much as I’d hope the transition is smooth, I don’t think it will be. In today’s 2025 labor market the employers have the power. They are looking to do more with less and asking candidates to jump through more hoops initially with pre-recorded interview answers will likely happen before companies do the same. And without the infrastructure to save and reuse those recordings, candidates will pay the price. Hopefully job boards will catch on to this quickly and provide the tools needed to make this efficient for everyone. (If you’re connected to them, be sure to let them know about this series of articles so we can accelerate this process.) While the end state will be better for everyone, the process to get there will likely be quite bumpy.
It’s critical to learn about corporate culture before you accept a job offer but it can be awkward to raise such questions. Learn what to ask and how to ask it to avoid landing yourself in a bad situation.
Investing just a few hours per year will help you focus and advance in your career.
Groups with a high barrier to entry and high trust are often the most valuable groups to join.
